Role, Workflow and Challenges

Transcription

1 1 Workshop on checks and management of agricultural land in IACS The actual use of imagery Image Processing at EFTAS Role, Workflow and Challenges Head of IT, Image Processing Unit CwRS

2 2 The Role 30 years of Remote Sensing at EFTAS First CwRS project in 1994 with a staff of < 10 people In 2018 EFTAS is dealing with 9 clients in DE and LU in the context of CwRS The zones 47 CwRS zones 12 zones with Image Processing only The staff 52 Interpreters for CAPI (shift work from 6am to 11pm) 4 controllers 4 shift supervisors 2 project coordinators 2 project managers 4 members of Image Processing Team

3 3 The Role The Methods 21 zones with VHR1 Sat + HR+1 + HR+2 15 zones with VHR1 Aerial + HR+1 + HR+2 5 zones with VHR1 Aerial + HRB1 + VHR2 Aerial + HR+1 3 zones with VHR1 Aerial + VHR2 Sat + HR+1 2 zones with VHR1 Aerial + VHR2 Sat + HR+1 + HR+2 2 zones with HR-1 + VHR1 Sat + HR+1 11 zones with VHR1 Aerial (sdop) + S2 Imagery

4 4 The Role The Methods Optimizing the image requirements workflow: We absolutely need VHR and HHR Imagery at the current state of CwRS

5 5 The Role The Imagery 28 VHR Sat windows 30 VHR Aerial windows 93 HHR windows 11 sdop windows (Imagery delivered by national surveying offices) 25 LPIS QA windows (approximately) The Time Frame results for crop diversification checks results for all other checks incl. summer crops 4 weeks delivery of processed Imagery to Administration after receipt

6 6 The Workflow The Processing Chain Download / Provision Import (Mosaicking of VHR Tiles) Ortho-Rectification / Aero-Triangulation Accuracy Checks LUT-Stretch 8Bit Conversion Cutlines and Mosaicking DEM-Masking Tiling Export

7 7 The Workflow

8 8 The Workflow The Results (Accuracies achieved for VHR Imagery in 2017)

9 9 The Workflow The Results (Accuracies achieved for HHR Imagery in 2017)

10 10 HHR [F2.] versus [F1.] EFTAS does not use [F2.] Imagery due to insufficient accuracy PAN-Sharpening has to be performed by EFTAS SPOT , zone PENZ. Left: [F1.] processed by EFTAS with DEM5, Right: [F2.] provided by Airbus DS with unknown DEM origin. Spot 6/7 Airbus DS (2017). Data received via the Joint Research Centre of the European Commission under FWC

11 11 HHR [F2.] versus [F1.] Optimizing the image requirements workflow: Why not delivering [F1.] as PSH instead of Bundle only?

12 12 Haze in VHR Imagery 10 of 29 VHR Acquisitions have been tagged with MeteoFlag due to Haze (34,5%) Haze-Tagging is not very comprehensible Left: Zone HUDE, WV , 0% CC. Right: Zone VETT, WV , 1,5% CC. Both Acquisitions with MeteoFlag (Haze) WorldView-2/4 data, European Space Imaging/DigitalGlobe, 2018, distributed by European Space Imaging. Data received via the JointResearch Centre of the European Commission under FWC and FWC

13 13 Haze in VHR Imagery Optimizing the image requirements workflow: Can Haze-Tagging be more standardised to make it more comprehensible and understandable to end-users?

14 14 Delivery of partials / fragmentation of VHR Imagery The delivery of Partials is not generally a problem, but it becomes more difficult and extends the Image Processing if partials are too small or of an exceptional shape Zone BELU, GE Blue Strips acquired W-E and green strips acquired N-S. Strip width-limitation leads to a very narrow Image part GeoEye-1 data, European Space Imaging/ DigitalGlobe, 2018, distributed by European Space Imaging. Data received via the Joint Research Centre of the European Commission under FWC and FWC

15 15 Delivery of partials / fragmentation of VHR Imagery Triangle-like partials Zone EGGE, GE Blue Strip acquired a few days later than the green ones, on which the north-western part had been cloudy. Unfortunately only the triangle has been provided. GeoEye-1 data, European Space Imaging/ DigitalGlobe, 2018, distributed by European Space Imaging. Data received via the Joint Research Centre of the European Commission under FWC and FWC

16 16 Delivery of partials / fragmentation of VHR Imagery Optimizing the image requirements workflow: Imagery could be processed easier and faster if partials were delivered as rectangles instead of exceptional shapes, like triangles

17 17 Delivery of partials / mix of strips with very diverging Elevation Angles Low ELA together with high ELA prevent a better pixel size A pixel size better than 0,5m had been possible if Strip 1 came with a higher ELA Actually a pixel of 0,5m is recommended in this case

18 18 Delivery of partials / mix of sensors with very diverging GSD, e.g. K3 and WV3, WV4 A sensor mix of WV and K3 is always critical and not favored Actually a pixel of 0,5m is recommended in this case, even if large parts of the Image will be of worse / coarser resolution All parts of a single strip have to be triangulated by using GCPs and Tie Points In this example 35 GCPs and at least 30 Tie Points Are neccessary

19 19 Delivery of partials / fragmentation of VHR Imagery Optimizing the image requirements workflow: Imagery could be processed more precise and more homogeneously with partials of the same or at least similar sensors

20 20 Not enough time to explain in detail thus just mentioning it: too much time between first and last partial acquisition chosen profiles others than [A1.] are often changed back to [A1.] by feasibility elevation angles lower than 50 may lead to pixel sizes too coarse OIR by FTP takes too much time IDQA: The Acquired area stated by Provider cannot be confirmed or rejected at that early stage of processing Zones definition: Shrinking of zone s size after ortho-rectification: Automatical extension of outer limits by ImageProvider to be continued

21 21 Thank you very much for your attention! contact: EFTAS Image Processing for CwRS name:, Head of IT, Image Processing Unit CwRS phone: Vilnius, EFTAS 28 th Fernerkundung 30 th of May Technologietransfer 2018 GmbH ı Oststr ı Münster ı Robert Germany Stein ı ı Tel.: +49 (0) ı Fax: +49 (0)